The registry case finding engine: an automated tool to identify cancer cases from unstructured, free-text pathology reports and clinical notes.

摘要

BACKGROUND: The American College of Surgeons mandates the maintenance of a cancer registry for hospitals seeking accreditation. At the University of Michigan Health System, more than 90% of all registry patients are identified by manual review, a method common to many institutions. We hypothesized that an automated computer system could accurately perform this time- and labor-intensive task. We created a tool to automatically scan free-text medical documents for terms relevant to cancer. STUDY DESIGN: We developed custom-made lists containing approximately 2,500 terms and phrases and 800 SNOMED codes. Text is processed by the Case Finding Engine (CaFE), and relevant terms are highlighted for review by a registrar and used to populate the registry database. We tested our system by comparing results from the CaFE to those by trained registrars who read through 2,200 pathology reports and marked relevant cases for the registry. The clinical documentation (eg, electronic chart notes) of an additional 476 patients was also reviewed by registrars and compared with the automated process by the CaFE. RESULTS: For pathology reports, the sensitivity for automated case identification was 100%, but specificity was 85.0%. For clinical documentation, sensitivity was 100% and specificity was 73.7%. Types of errors made by the CaFE were categorized to direct additional improvements. Use of the CaFE has resulted in a considerable increase in the number of cases added to the registry each month. CONCLUSIONS: The system has been well accepted by our registrars. CaFE can improve the accuracy and efficiency of tumor registry personnel and helps ensure that cancer cases are not overlooked.